Restricing themes and thesaurus extensions

Question

Hi Gurus,
Couple of quick q's:
1. How can we restrict the themes?
We usually generate themes by
Step1: creating a text index using indextype is ctxsys.context parameters ('lexer themes_only');
Step2: select token_text, token_count from SCHEMA.DR$TABLE_INDEX0$I
order by token_count desc;
This usually works great for producing both the keywords and themes depending on the context parameters we chose.
However some times we end up with themes such as:
BUSINESS AND ECONOMICS
ABSTRACT IDEAS AND CONCEPTS
GENERAL COMMERCE
COMMERCE AND TRADE
HUMAN LIFE AND ACTIVITY
SOCIAL ATTITUDE, CUSTOM
which probably make some sense but in general are way too general. So is there a way we can restrict themes to the leaves in the hierarchy and not consider the branches of the knowledge base? (I mean - can we obtain only more specific themes?).
2. We followed Barbara's advice on using thesaurus for including search synonyms for terms using the discussion:
Re: Loading an Oracle Thesaurus:
The discussion helps us to create a thesaurus and use it in text queries using contains and SYN.
However, if we are trying to use for text clustering, or generating themes/gist how can use the created thesaurus?
I searched the forums but was not successful in finding how one can use the thesaurus for clustering or themes/gist.
Please advice

Answer 1

Hi,
1. - solved...
2. We know how to perform clustering in cases where we use default thesaurus. But I am not sure how to specify the thesaurus we created while performing clustering or summarization as these both processes depend on the themes that are generated.
For example:
In the below code for clustering where can we overide/append the default thesaurus?
/* set the preference */
exec ctx_ddl.drop_preference('my_cluster');
exec ctx_ddl.create_preference('my_cluster','KMEAN_CLUSTERING');
exec ctx_ddl.set_attribute('my_cluster','CLUSTER_NUM','3');
/* do the clustering */
exec ctx_output.start_log('my_log');
exec ctx_cls.clustering('collectionx','id','restab','clusters','my_cluster');
exec ctx_output.end_log;
After performing clustering, we see that the synonyms for say customer - customers, cust, cst - are not present in similar clusters. How can I make sure the relations specified in the thesaurus are used by the clustering algorithm?
Thanks for your quick response,
Please guide.